Update app.py

app.py

@@ -74,12 +74,16 @@ def get_model(temperature, top_p, repetition_penalty):
 def generate_chunked_response(model, prompt, max_tokens=1000, max_chunks=5):
     full_response = ""
     for i in range(max_chunks):
-        chunk = model(prompt + full_response, max_new_tokens=max_tokens)
-        chunk = chunk.strip()
-        if chunk.endswith((".", "!", "?")):
+        try:
+            chunk = model(prompt + full_response, max_new_tokens=max_tokens)
+            chunk = chunk.strip()
+            if chunk.endswith((".", "!", "?")):
+                full_response += chunk
+                break
             full_response += chunk
+        except Exception as e:
+            print(f"Error in generate_chunked_response: {e}")
             break
-        full_response += chunk
     return full_response.strip()
 
 def extract_text_from_webpage(html):
@@ -209,7 +213,12 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search):
     else:
         database = None
 
-    if web_search:
+    max_attempts = 3
+    context_reduction_factor = 0.7
+
+    for attempt in range(max_attempts):
+        try:
+            if web_search:
         original_query = question
         rephrased_query = rephrase_for_search(original_query, model)
         print(f"Original query: {original_query}")
@@ -242,6 +251,7 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search):
         """
         prompt_val = ChatPromptTemplate.from_template(prompt_template)
         formatted_prompt = prompt_val.format(context=context_str, original_question=question, rephrased_query=rephrased_query)
+    
     else:
         if database is None:
             return "No documents available. Please upload documents or enable web search to answer questions."
@@ -250,6 +260,11 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search):
         relevant_docs = retriever.get_relevant_documents(question)
         context_str = "\n".join([doc.page_content for doc in relevant_docs])
 
+        # Reduce context if we're not on the first attempt
+        if attempt > 0:
+            words = context_str.split()
+            context_str = " ".join(words[:int(len(words) * context_reduction_factor)])
+
         prompt_template = """
         Answer the question based on the following context:
         Context:
@@ -287,6 +302,15 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search):
 
     return answer
 
+        except Exception as e:
+            print(f"Error in ask_question (attempt {attempt + 1}): {e}")
+            if "Input validation error" in str(e) and attempt < max_attempts - 1:
+                print(f"Reducing context length for next attempt")
+            elif attempt == max_attempts - 1:
+                return f"I apologize, but I'm having trouble processing your question due to its length or complexity. Could you please try rephrasing it more concisely?"
+
+    return "An unexpected error occurred. Please try again later."
+
 # Gradio interface
 with gr.Blocks() as demo:
     gr.Markdown("# Chat with your PDF documents and Web Search")